Constraint

0/6

Topic 05 · Necessary But Not Sufficient

The tool worked. The system did not.

AI assistants, faster CI, a new platform, another observability stack. The technology is real. Throughput stays flat when the rules, the release of work, and the measures were built for the previous constraint. Installing the tool on top of those rules automates the waste, or moves the queue.

After Haystack. Six sections. The constraint and the rope are the ones from Flow.

01

Necessary, not sufficient

Goldratt’s argument about ERP, applied to anything sold as a transformation of delivery. The tool has to work. That is the necessary part.

A system only produces a lasting gain when the organisation also changes the policies and the measurements that were built around the old limitation. Two out of three is still not enough.

02

Three conditions

The technology is available. The process is redesigned around whatever the tool just released. The measures reward Throughput, not the local activity the tool made cheap.

  • In place

    Technology

  • Missing

    Process

  • Missing

    Measures

Throughput

Still not sufficient

All three arrows have to hold. Any two are necessary and still not enough.

Necessary but not sufficient. Still missing a process change and a measurement change.

03

The queue moves

More code, same cycle time, leadership pleased by the activity. The constraint left coding. Nobody followed it.

Coding queue12
Review queue2

Throughput 4 used outcomes / week

Coding is the long queue. Review is waiting on code.

Review is unchanged. More work is started because coding feels faster. Success is counted in commits. The rope does not move.

04

Old rules

Heavy review, large batches, utilisation, detailed estimates. Each one protected a real limitation. When the limitation shrinks, the rule can become the constraint.

  • Every change needs the same heavy review

  • Plan a large batch before anyone starts

  • Utilisation targets

  • A person still has to approve production

05

Make it sufficient

Treat the rollout as a change to technology, process, and measurement. Then look again. The constraint will not be where the vendor’s diagram left it.

  1. Name the constraint before you buy.
  2. If the technology works perfectly, where does the constraint move?
  3. Subordinate to that new place. WIP, pull, protection.
  4. Change the measures so local activity is no longer the win.
  5. After it is in, re-identify. Do not keep the old policy out of inertia.

06

The next tool, before you sign

If you cannot name the process change and the measurement change, the tool is likely to be necessary and not sufficient.

  1. The constraint it is supposed to address.
  2. The process and the measure that have to change for the gain to appear.
  3. If those sentences are empty, stop.
  4. After adoption: did valuable outcomes rise, or did the queue move?
  5. Remove the policy the technology has made obsolete. Keep the one that still protects Throughput.

This week

0/5

Cheat sheet

The claim
Technology is necessary. It is not sufficient.
Three together
The tool works. The process changes. The measures change. Any two still fail.
Install only
Old policies stay. Waste is automated, or the queue moves and looks like progress.
Where it moves
Faster coding usually exposes review, test, a product decision, or deploy.
Fossils
Many rules exist because of a limitation the tool just removed.
After
Re-identify the constraint. Do not leave the old policy in place out of habit.

Check yourself

5 questions. No timer.

1A tool raises Throughput for good when…
2AI coding tools often leave Throughput flat because…
3Adding build agents when builds are not the wait…
4An old policy becomes a fossil when…
5The number to watch after a rollout is…